A minimal PBPK model to unveil anti-tuberculosis drug distribution in pulmonary lesions: a blueprint for improving efficacy screenings.
Anna Fochesato (1,2), Federico Reali (1), Roberto Visintainer (1), Shayne Watson (3), Micha Levi (3), Véronique Dartois (4), Luca Marchetti (1,5)
(1) Fondazione The Microsoft Research – University of Trento Centre for Computational and Systems Biology (COSBI), Italy, (2) University of Trento, Department of Mathematics, Italy, (3) Gates Medical Research Institute, USA, (4) Hackensack Meridian Health, USA, (5) University of Trento, Department of Cellular, Computational and Integrative Biology (CIBIO), Italy.
Objectives: With a human toll of 1.6 million lives in 2021, tuberculosis (TB) is among the most aggressive infectious diseases worldwide, continually exasperated by the emergence of novel resistant strains [1]. As potent and highly penetrant drugs are necessary for hard-to-treat TB phenotypes [2], quantifying the exposure of anti-TB candidates at the disease site-of-action is pivotal for dosage optimization and efficacy assessment. Given the invasiveness of lung biopsies in clinics, animal models are valuable surrogates to study the in vivo pharmacokinetics (PK) of these agents in TB granuloma and caseum and to bridge the gap with clinics. The work aims at developing a flexible and lesion-centric PBPK-based platform in rabbits that can support a large pool of anti-TB drugs and their active metabolites. The model intends to provide a novel PK description of the intra-pulmonary environment and, coupled with a mouse PK work, to pose a foundation for mechanistic site-of-action translations.
Methods: A minimal physiologically-based pharmacokinetic (mPBPK) model was designed to describe anti-TB compound disposition in TB-infected rabbits with a resolution down to the necrotic caseum. TB-unessential organs were lumped together to streamline the model complexity while retaining reliable physiological features. THP-1 uptake and free fraction in caseum assays were leveraged to inform drug transport modeling within TB lesions. In silico Rodgers and Rowland equations [3] were used to compute tissue-to-plasma partition coefficients, while drug physicochemical properties and rabbit anatomical attributes were retrieved from the literature [4]. PK measurements for eleven anti-TB compounds sampled in plasma, uninvolved lung, cellular lesion, and caseum of New Zealand White rabbits were used to benchmark the minimal design. The model was implemented in Matlab R2019b and simulated via the ode15s function. Parameter estimation was performed through the Covariance Matrix Adaptation – Evolution Strategy (CMA-ES), a state-of-the-art genetic optimization algorithm [5].
Results: A structural identifiability analysis carried out on the minimal model ensured the well-posedness of the estimation task in our PK observable setting. Visual predictive checks for a 1500-rabbit virtual population guaranteed a good agreement between model predictions and observed PK profiles in plasma, lungs, and TB lesions for all the considered drugs (rifampicin, rifapentine, isoniazid, pyrazinamide, moxifloxacin, pretomanid, delamanid, sutezolid, bedaquiline, clofazimine, and a novel carbostyril derivative). Simulated PK profiles were compared to gold standard MIC50 (minimum inhibitory concentration to inhibit bacterial growth by 50% bacteria) and novel ex vivo casMBC90 (caseum minimum bactericidal concentration to kill 90% bacteria) potency metrics to evaluate lesion coverage [6]. Leveraging an in-house BALB/c mouse mPBPK work, model-based relationships were drawn to explain species dependencies and inform model scalability.
Conclusions: The framework we developed requires low-level prior information to accurately reconstruct PK dynamics in hard-to-treat TB sites for several anti-TB agents, providing an easy-to-use and fast virtual lab for large-size what-if investigations. Model simulations questioned current experimental assumptions showing a mismatch between drug levels in the systemic circulation and in TB lesions for certain anti-TB drugs. Informed by a mouse-to-rabbit scaling analysis, the work acts as a pilot study for lesion-centric and mechanistic-based PK projections in clinical trials.
References:
[1] World Health Organization (2022), Global tuberculosis report 2022.
[2] Imperial et al., Precision-Enhancing Risk Stratification Tools for Selecting Optimal Treatment Durations in Tuberculosis Clinical Trials, American Journal of Respiratory and Critical Care Medicine, Volume 204, Issue 9, 2021.
[3] Rodgers et al., Physiologically based pharmacokinetic modelling 2: predicting the tissue distribution of acids, very weak bases, neutrals and zwitterions, Journal of Pharmaceutical Sciences 95(6):1238-57, 2006.
[4] Mavroudis et al., Development and validation of a physiology-based model for the prediction of pharmacokinetics/toxicokinetics in rabbits, Plos One, 2021.
[5] Hansen Nikolaus, The CMA Evolution Strategy: A Comparing Review; in Lozano et al., Towards a New Evolutionary Computation. Studies in Fuzziness and Soft Computing, vol 192, pp 75-102, Springer, 2006.
[6] Sarathy et al., Extreme Drug Tolerance of Mycobacterium tuberculosis in Caseum, Antimicrobial Agents and Chemotherapy, Vol. 62, No.2, 2018.